Metrology for fluxes: eddy covariance measurement uncertainty

The uncertainty evaluation of eddy covariance flux measurements has been thoroughly developed in the last two decades. However, the various methods proposed are not yet fully compliant with the internationally accepted metrological guidelines, e.g. those indicated in the Guide to expression of uncertainty in measurement and related supplements issued by the Joint Committee for Guides in Metrology and internationally adopted as reference in metrology. Scope of this presentation is to implement the formal methodology for the determination of a combined standard uncertainty for the estimated fluxes through the law of propagation of uncertainty, assuming independent variables. Compared to previous methods, this approach considers the complete flux equation, including the coordinate rotations and the physical conversions and, most importantly, provides an easy to implement analytical tool to quantify the individual contributions to the full measurement uncertainty arising from all the variables actually included in the calculation (turbulent wind components, scalar of interest, air temperature and pressure). The linear method adopted for uncertainty propagation has been also validated through a Monte Carlo simulation, which is the gold standard for propagating probability distributions. The methodology has been applied to a full year of carbon dioxide fluxes measured in the San Rossore 2 ICOS Ecosystem Station, a Mediterranean forest, but it is valid for most of the common eddy covariance systems, being based on theoretical principles. The median of the estimated relative uncertainty of the flux over the considered year is 13.5%, assuming an instrumental uncertainty of 30 Pa for the barometer, 0.5 °C for the thermometer, 4 ppm for the CO₂ analyzer and 0.4 m/s for the three components of the sonic anemometer. The main uncertainty contributions come from the analyzer and the vertical component of the anemometer, with medians of the evaluated relative uncertainties equal to 11.9% and 3.25%, respectively. Preliminary results suggest that the method is robust and confirm expectations about the relative contribution of the different instruments used for flux determination, but at the same time constitute a tool for a sounder metrological assessment of all eddy covariance measurements and applications.